Abstract
Hookups can result in sexual assault when men do not listen to requests from women to stop. It is thus important to identify factors that influence men’s decisions to override direct refusals in these situations. Presently, we administered first-person vignettes depicting a prototypical hookup wherein the woman refuses the man’s attempt to escalate intimacy. Using a national sample of emerging adult men (N = 420), we found that they on average did not completely rule out coercive or forcible tactics, but those elevated on rape myth acceptance, hypermasculinity, and psychopathy were uniquely at risk of assault when controlling for several other traits known to correlate with rape. Participants also reported being likelier to use coercive sexual practices when refusals occurred at higher levels of sexual intimacy already attained. Notably, diagnostic analyses revealed that a subset of men had a disproportionate influence on the regression estimates, and that these men were not only elevated across a range of assault-relevant traits, but also endorsed higher likelihoods of using coercion and force in the face of female sexual refusal. Although removal of these cases did not substantively alter the results, exploratory analyses revealed that these individuals responded differently to situational factors in ways that suggested sexual opportunism. Theoretical and practical implications of these findings are discussed.
Introduction
Male-against-female sexual assault is extremely prevalent among emerging adults (Mumford et al., 2020). Through various channels, men in this demographic are exposed ideologically to hookup culture (Olmstead et al., 2019; Pugh & Becker, 2018), which privileges male sexual conquest and reinforces their dominance over women. This occurs at a time when peer norms are influential and sexual practices solidify. Successful intervention at this juncture may have both short- and long-term impacts on rates of sexual assault. Successful intervention, however, requires an understanding of the factors that contribute to sexual assaults and influence emerging adult men’s specific actions in sexual situations.
One way to examine men’s decision making during sexual assaults is to employ measures of rape proclivity, or self-reports of the likelihood that one would commit a hypothetical rape. Rates of endorsement for hypothetical rape are alarmingly high among emerging adult men (Palmer et al., 2020), and endorsement of a hypothetical rape is empirically linked to actual sexual misconduct (e.g., Bohner et al., 2005). As such, measures of rape proclivity may provide a reasonable proxy for a given man’s general risk for perpetration, and thus be a useful metric for identifying dispositional factors that link to sexual assault. For instance, there is an established a link between rape myth acceptance and rape proclivity (e.g., Chiroro et al., 2004). Presently, however, research on rape proclivity has focused on only a small number of dispositional variables vis-à-vis the set of characteristics implicated in sexual assault perpetration. These dispositions are highly intercorrelated but are often examined in isolation. It is thus unclear if any effects found when studying a given variable are truly specific to that variable’s domain (e.g., sexist beliefs). They could instead reflect shared variance with some other related construct (e.g., traditional gender role beliefs) or represent a broader latent trait (e.g., toxic masculinity), as some theories would suggest (e.g., the Confluence Model; Malamuth et al., 1995).
There is also a dearth of research on situational factors associated with rape proclivity in particular scenarios. Although research demonstrates that specific situational factors associate with men’s misperceptions of consent, few studies have examined how situational manipulations influence rape proclivity per se. This distinction is important because the misconception of consent is only one path by which context may increase assault risk, as men also may view specific situational factors opportunistically (e.g., female alcohol use; see Recalde-Esnoz et al., 2023). Measures of rape proclivity also allow researchers to investigate sexual decision making within circumscribed situations, which is important given the heterogeneous nature of sexual assaults. Research focused on high-risk contexts may elucidate the most relevant determinants of sexual assaults in situations where they are likeliest to occur (e.g., Lindo et al., 2018). Hookups involving cisgender heterosexual men are particularly risky for women (Mellins et al., 2017). Hookups can include various sexual activities between partners who may or may not be strangers, but between whom commitment is not presumed (Garcia et al., 2012). Although hookups, by definition, involve some consensual sexual activity, evidence suggests that women, on average, do not want to escalate the level of sexual intimacy during hookups as far as do men (Olmstead et al., 2017). This discrepancy contributes to situations wherein a male pursues sexual intimacy goals that exceed those of his female hookup partner.
Although in some cases women do not actively refuse unwanted sexual advances (e.g., due to fear or pressure; Pugh & Becker, 2018), research suggests that they frequently employ a variety of explicit and implicit communication strategies to refuse men’s sexual initiations (Richards et al., 2022). Importantly, men appear to readily understand the meaning of such cues across different modes of communication (Lofgreen et al., 2017), even when subtly conveyed (O’Byrne et al., 2006), and often respond appropriately (Jozkowski et al., 2017). Given the high rate of assaults occurring during hookups, however, it seems likely that across many of these situations some men opt for a coercive or forcible approach to achieve their sexual goals despite understanding the message of refusal. Currently, little is known about what factors are most influential to emerging adult men’s decision-making in sexual situations broadly, let alone the choice to engage in a deliberate assault against a hookup partner who overtly refuses increased intimacy. Such research can help explain the heightened risk of sexual assault for women in heterosexual hookups, as well as identify interventions to mitigate this risk.
Dispositional Correlates of Sexual Assault Risk
Men inclined towards sexual misconduct tend to report similar constellations of traits, often proportional with the extremity of their behaviors (Swartout et al., 2015b). Although some attributes have primarily been studied in forensic populations (e.g., interest in pedophilic crimes; Stephens et al., 2021), other traits – such as the acceptance of rape myths – have been identified as common risk factors across a variety of populations (Yapp & Quayle, 2018), including undetected offenders. Additional well-researched risk factors for perpetration include hypermasculinity (i.e., rigid adherence to male-coded traits such as dominance; Vechiu, 2019), psychopathic traits (i.e., a personality formation featuring callous disregard for others; Mouilso & Calhoun, 2013), hostile sexism (i.e., the view that relations between the sexes are inherently antagonistic; Glick & Fiske, 1996), and low levels of trait empathy (i.e., the ability to understand and feel what another person feels; DeGue & DiLillo, 2004).
Research suggests that these dispositional risk factors can operate globally (e.g., the hostile masculinity construct of the Confluence Model; Malamuth et al., 1995), raising the risk of sexual misconduct perpetration in general, or more narrowly through mechanisms theorized to be specific to a given personality or attitude trait (e.g., hostile sexism; Abrams et al., 2003). These complex processes have led to a bifurcation in the literature, with some researchers focusing on risk factors arising from intercorrelated and synergistically operating dispositions (Woerner et al., 2018), while other researchers focus on single traits as risk factors in isolation (e.g., rape myth acceptance; Bohner et al., 2005). Both of these approaches contribute to the literature, but render it unclear whether any observed effects are domain-specific or reflect a single broader trait. Although higher-order risk factors are certainly relevant to perpetration, identifying whether individual variables have separable unique effects is also important, as this could point to specific mechanisms underlying sexual misconduct across different situations, as well as rule out proposed mechanisms tied to variables that lack unique associations with sexual assault.
There is reason to believe that different predisposing factors do have their own unique sets of mechanisms that link them with sexual assaults during hookups, specifically those that occur in the face of overt refusals. Men subscribing to rape myths may believe that force is justified if a man interprets a woman’s behavior as indicating interest in sexual activity, even if she also directly refuses, or they may conclude that a woman’s resistance belies a genuine desire for domination (Lynch et al., 2017). In contrast, men hostile towards women may characterize those who refuse as deceitful and knowingly assault her as retribution. Men high in psychopathy may accurately infer the woman’s lack of consent, but are prone to violating the rights of others, whereas those elevated in hypermasculinity may utilize aggressive behavior to enforce traditional gender roles (e.g., Reidy et al., 2009). Low empathy may remove barriers to offending by reducing the ability to recognize the distress of the other person (DeGue & DiLillo, 2004).
There also exist other variables with potential ties to sexual assault that are not as well-researched, but nevertheless may carry unique influences, or otherwise explain the association of other candidate predictors. Social dominance ideologies, for instance, promote status discrepancies between groups and aggression toward subordinates (Ho et al., 2012). Similarly, individuals high in right-wing authoritarianism are characterized by acceptance of aggression endorsed by authority figures and adherence to conventional values. These attitudes are closely related with prejudice towards marginalized groups, including women, and social dominance is particularly linked to tolerance of sexual harassment and aggressive responses to romantic rejection (e.g., Kelly et al., 2015). More research is needed to parse apart these relations, as both social dominance orientation and right-wing authoritarianism correlate positively with each other (particularly in student samples, Roccato & Ricolfi, 2005), and carry varied empirical and theoretical links with psychopathy (e.g., Glenn et al., 2017), traditional masculinity (e.g., Goodnight et al., 2014), rape myths (e.g., Kelly et al., 2015), and hostile sexism (Feather & McKee, 2012). Benevolent sexist attitudes may also uniquely associate with sexual assault risk. These refer to a set of patronizing or even infantilizing beliefs that frame women as inherently demure and fragile (Glick & Fiske, 1996), and function to place men in control of decision-making (sexual and otherwise) by positioning women as passive recipients of sexual attention from confident, domineering men. In the realm of sexual assault, benevolent sexism uniquely relates to the de-agentification of women and their ability to consent (see Fraser, 2015).
Sexual assault risk is also associated with sociosexual orientation, which refers to the tendency towards uncommitted and impersonal sexual relationships. The link with sexually aggressive behaviors is particularly strong for men engaging in frequent casual sex (Malamuth et al., 1995), and there is evidence that phenotypes for high sociosexuality and sexual coercion have a common genetic basis (Westerlund et al., 2010). Moreover, Yost and Zurbriggen (2006) found that sociosexual orientation associated with rape myths and hypermasculine traits (e.g., power motives), which may account for their observed association with sexual assault, as it has been shown to do for psychopathy (Mouilso & Calhoun, 2012).
Situational Influences on Assault Risk
Numerous situational characteristics may affect a man’s use of sexual assault tactics, irrespective of his personality or beliefs. One key factor appears to be the prior sexual behavior of the parties involved, otherwise known as sexual precedence. As the level of intimacy in an ongoing sexual interaction increases, men may begin to take for granted the woman’s consent to intercourse and, correspondingly, may expend less effort to assess her willingness (see Lofgreen et al., 2017), or dismiss refusals as belying the woman’s genuine sexual intentions. The extent of prior sexual interactions between two individuals is also perceived to decrease the importance of explicitly confirming consent (Beres, 2014). Likewise, women known to have more extensive sexual histories are viewed by third parties with more skepticism and blame when victimized, and are considered less deserving of empathy than those with less extensive histories (e.g., Gillmor et al., 2014). Similar effects have been observed for women adopting stereotyped indicators of sexual availability (e.g., revealing attire; Moor, 2010). Insofar as men consider third-party blame attributions, or simply have internalized rape myths, they may feel safer or more justified in assaulting women disparaged by society. Indeed, men seem to be more likely to commit a sexual assault if they anticipate fewer consequences (Stotzer & MacCartney, 2016).
Victim blame also occurs when women consume alcohol (e.g., Sims et al., 2007), which is viewed as a violation of appropriate gender roles, and thereby can elicit blame for victimization. By comparison, intoxicated men may be granted a culturally-condoned “time-out” from the constraints of acceptable behavior (Peralta, 2010). The reallocation of blame associated with alcohol use may influence men’s sexual decision-making directly by reducing the perceived costs of perpetration, while alcohol use also potentially increases men’s perception that a woman is sexually interested (e.g., Crane et al., 2016). In summary, the presence of alcohol during a sexual assault serves both to exonerate men and castigate women for the same consumption.
These findings together suggest that sexual precedence, perceived sexual availability, and alcohol use by one or both partners are associated with sexual assault and frequently features of heterosexual hookups. However, whether and how these factors impact men’s behavior following an overt sexual refusal during a hookup is not entirely clear. Though a direct “no” might undercut the impact of situational influences, Lofgreen et al. (2017) found that situational factors – specifically sexual precedence – increased men’s perceptions of consent across myriad scenarios, even those involving refusals. This may implicate the internalization of predominant sexual scripts, which can lead even well-intentioned men to dismiss a sexual refusal, or may influence men’s sexual decisions insofar as they affect the risk of consequences. Considered overall, it is plausible that situational factors implicated in sexual assault may more broadly impact decision-making in the specific context of sexual refusals during hookups, but actual confirmation of these effects in this context awaits investigation.
The Current Study
We used a vignette paradigm to examine the associations between a range of dispositional and situational variables on men’s proclivity towards assault in response to a sexual refusal by women during prototypic hookup scenarios. We describe our outcome as a proclivity towards sexual assault rather than rape proclivity, as we focused on a variety of hookup scenarios which could involve a wider range of non-consensual sexual acts than what meets a legal or academic definition of rape. Our aim was to identify, in these scenarios, the dispositional and situational factors that uniquely incline the “average” emerging adult man to violate the “no means no” standard of consent or, in other words, to attempt to coerce or advance sexual intimacy in response to explicit communication from the woman to stop. Although how a man responds in a hypothetical scenario may differ than how he would in vivo, the former specifically targets the man’s sexual calculus; to wit, how he thinks he would behave in a given situation.
We evaluated men’s reported likelihood of using coercive and/or forcible assault tactics. Extant measures of rape proclivity frequently use completed narratives that assess men’s inclination towards the most violent or extreme forms of rape (e.g., Malamuth et al., 1980). This is important to evaluate, but may be more subject to reporting biases. Moreover, sexual assaults are a broader phenomenon than rape (e.g., unwanted touching), and men who would not use extreme physical force may still be prone to sexual assaults by other means. Finally, sexual assaults infrequently conform to the classic rape script (Baum & Klaus, 2005). A more typical scenario is that of an initially consensual hookup wherein the woman attempts to halt the escalation of intimacy, and the man tries to override this refusal using behaviors designed to entice or pressure her (e.g., repeated requests). This unfolding aligns with sexual scripts, central to which is the notion that a woman’s role is to conceal her ultimate desire for sex and that it is the man’s role to ‘work out a yes’ (Osman, 2013). It is also important to evaluate what underlies a man’s willingness to press his physical advantage in various ways; that is, what increases the likelihood he would become physically aggressive in the face of refusal. Though coercive tactics in response to female resistance may reflect adherence to a predominantly-held script for an ostensibly consensual sexual interaction, the use of force to ensure sexual compliance may relate to a different set of mechanisms representing a disregard for female agency.
Predictors were identified through our literature review and selected on the basis of their well-established associations with sexual assault, victim blaming, and men’s misperceptions of consent. At the dispositional level, we surveyed rape myth acceptance, hypermasculinity, hostile and benevolent sexism, psychopathic personality features, empathy, sociosexual orientation, social dominance orientation, and right-wing authoritarian beliefs. Our aim was to identify which of these characteristics, if any, associate uniquely with men’s use of sexual assault tactics in hookups wherein the female partner refuses his attempt to escalate intimacy. We also manipulated the woman’s attire, sexual history, prior sexual intercourse between the parties, the level of intimacy in the interaction, and alcohol consumption. There is reason to believe that any and all of these variables may increase men’s sexual assault proclivity, and it is also possible that men use aspects of the situation opportunistically. Highlighting the factors that reliably influence men’s use of assault tactics across situations is important, as they may be most frequently implicated in tipping the decisional balance towards assault in the real world.
Methods
Participants
Estimating sample size was complicated by the multilevel design (see Maas & Hox, 2005), the lack of prior data to estimate nesting effects, and that our predictors of interest ranged in effect size and were to varying degrees intercorrelated. We thus took a conservative approach and estimated the sample size required to detect a small effect from an individual predictor in an OLS regression model with standard alpha (.05) and power (.80). We then used this estimate (N = 395) as a minimum number of independent observations required, and sought to oversample this number assuming that data from some number of respondents would be deemed unusable. Our initial sample comprised 598 heterosexual men. Of these, 99 did not provide data on the primary outcome variables and were omitted. Several additional steps were taken to identify insufficient effort responders, with 79 individuals (15.8%) meeting at least one of our criteria: 25 individuals (5%) had improbably short completion times, 33 (6.6%) had zero response variance across reverse scored items, and 24 individuals (4.8%) were multivariate outliers. These data were removed from further analysis, yielding an ultimate sample size of 420 participants.
The average age was 21.29 (SD = 2.38; range = 18–25) and the sample primarily self-identified as Caucasian (n = 201, 47.9%) and Christian (n = 196, 46.6%). All participants resided in the United States, with the majority currently residing in the Northeast (n = 139, 33.1%), followed by Southeast (n = 100, 23.8%), Midwest (n = 59, 14.0%), Southwest (n = 59, 14.0%), West (n = 51, 12.1%), and other (n = 4, 1%), with eight men (1.9%) not reporting. A majority reported that their relationship status was single (n = 268, 63.8%) but having had sexual intercourse at least once with a woman (n = 309, 73.6%), with the modal number of lifetime sexual partners being between 2–3 (n = 81, 19.2%). Most participants were enrolled in post-secondary education (n = 270, 64.3%), with 14 (3.4%) not reporting their current educational status. In terms of completed education, 9 (2.1%) individuals reported having less than a high school education, whereas the remaining sample endorsed completion of a high school degree (n = 200, 47.6%), some college coursework (n = 98, 23.3%), a two-year (n = 39, 9.3%), a four-year (n = 49, 11.7%), a master’s degree (n = 14, 3.3%), or a doctoral-level degree (n = 8, 1.9%).
Measures
Ambivalent Sexism
We used the Ambivalent Sexism Inventory (ASI; Glick & Fiske, 1996), which contains two subscales that respectively measure positive (i.e., benevolent) and negative (i.e., hostile) forms of sexist beliefs. The Hostile Sexism (HS) subscale measures negative stereotypes and attitudes that characterize women as antagonistic towards men (e.g., “Women exaggerate problems they have at work.”), whereas Benevolent Sexism (BS) uses items capturing stereotypically idealized views of women as pure and in need of male protection (e.g., “Women should be cherished and protected by men.”). Participants responded to the 11 items on each scale using a 6-point Likert-type scale (1 = Strongly Disagree; 6 = Strongly Agree). Items were reverse scored as appropriate and averaged within each subscale, with higher scores indicating more sexist attitudes. ASI subscales demonstrate acceptable reliability and validity (Glick & Fiske, 1996), and are often examined as standalone measures in research on sexual assault and rape proclivity (e.g., Abrams et al., 2003). Items of the BS and HS scales were intermingled in the original order validated by Glick and Fiske. The sample average was 3.46 (SD = .93; α = .83) and 3.61 (SD = .81; α = 0.74) for the HS and BS scales, respectively.
Empathy
We used two subscales of the Interpersonal Reactivity Index (IRI; Davis, 1980). The Perspective Taking subscale includes seven items assessing participants’ tendency to adopt and/or consider others’ perspectives. The Empathic Concern subscale includes seven items assessing participants’ feelings of sympathy for others’ suffering. The IRI has acceptable psychometric properties, and has been used in research on sexual aggression (e.g., DeGue et al., 2010). Items were rated on a Likert-type scale (1 = Not at all like me; 5 = Just like me), reverse scored as appropriate, and averaged across both subscales. Higher scores equate to endorsement of greater levels of empathy. The sample mean was 3.53 (SD = .53; α = .73).
Hypermasculinity
Participants responded to the Male Role Attitudes Scale (MRAS; Pleck et al., 1993) and the Adolescent Masculinity Ideology in Relationships Scale (AMIRS; Chu et al., 2005) to assess adherence to stereotyped masculinity. These two measures capture different aspects of hypermasculinity. Items on the MRAS focus on maintaining true privilege and power in relationships, as compared with the AMIRS, which emphasizes projecting a convincing semblance of this status. The MRAS consists of eight items assessing individuals’ endorsement of norms related to status, toughness, and anti-femininity using a 4-point Likert-type scale (1 = Strongly disagree, 4 = Strongly agree). The AMIRS has 12 items rated on a 6-point Likert-type scale (1 = Strongly disagree, 6 = Strongly agree) that emphasize emotional invulnerability and heterosexual dominance. Items were reverse scored and averaged so that higher scores represent stronger alignment with hegemonic male role ideology. Sample averages were 2.57 (SD = .59; α = .78) for the MRAS and 2.79 (SD = .88; α = .84) for the AMIRS.
Psychopathy
The Levenson Self-Report Psychopathy Scale (SRPS, Levenson et al., 1995) assesses antisocial tendencies with 26 items rated on a 4-point Likert-type scale (1 = Strongly disagree; 4 = Strongly agree). The scale assesses primary and secondary psychopathy traits. Primary psychopathy reflects an idiopathic selfishness and callous manipulation of others, whereas secondary psychopathy reflects impulsive and self-defeating decision-making. The sample mean for the LSRP was 2.29 (SD = .47; α = .88).
Rape Myth Acceptance
The Illinois Rape Myth Acceptance Scale – Short Form (IRMA-SF; Payne et al., 1999) contains 20 items measuring attitudes frequently invoked to minimize or excuse rape. Participants responded on a 6-point Likert-type scale (1 = Strongly Disagree; 6 = Strongly Agree), with higher scores indicating greater agreement with rape myths. This measure is often utilized in research on sexual aggression and has adequate construct validity (Suarez & Gadalla, 2010). Responses were averaged across the scale. Higher scores equate to greater levels of rape myth endorsement. The coefficient alpha was .93 (M = 2.65, SD = 1.25).
Right-Wing Authoritarianism
The revised Right-Wing Authoritarianism Scale (RWA, Altemeyer, 2007) measures attitudes towards conservative authoritarianism. Participants responded to the 22 items on a 9-point scale (1 = Strongly Disagree, 9 = Strongly Agree). Items assess the degree to which individuals endorse traditional values espoused by authorities, defer to authority figures, and support authority-sanctioned aggression towards outgroups. Negatively-phrased items were reverse scored. We averaged across items so that higher scores represent stronger endorsement of authoritarianism. The sample average was 4.47 (SD = 1.24; α = .89).
Social Desirability
A 13-item version of the Crowne-Marlowe Social Desirability Scale (CMSDS; Crowne & Marlowe, 1960) assessed participants’ propensity towards positive impression management, and has been used in vignette studies to control for impression management (e.g., Angelone et al., 2007). The CMSDS contains true-false items (1 = True; 2 = False) assessing whether participants acknowledge taking part in common but socially undesirable behaviors. Items were keyed and averaged so that higher scores equate to greater concerns about impression management. A participant denying all socially desirable statements would have the minimum score, 1, while a participant endorsing all statements would reach the maximum of 2. The coefficient alpha (.48) was substandard (M = 1.48, SD = .18).
Social Dominance Orientation
The revised Social Dominance Orientation scale (SDO7; Ho et al., 2012) captures a preference for the hierarchical distribution of power based on group identification. Participants responded to the 16 items on a 7-point Likert-type scale (1 = Strongly Oppose, 7 = Strongly Agree), with higher average scores representing stronger endorsement of hierarchical dominance in society. The sample average was 3.24 (SD = 1.07; α = .89).
Sociosexuality
The Revised Sociosexual Orientation Inventory (SOI-R; Penke, 2011) assesses participants’ endorsement of uncommitted sexual relationships. The first three items of the SOI-R have participants report how many sexual partners they have had for three different casual sexual activities. For each of these three items, respondents use a 9-point ordinal scale with options ranging from 0 to 20 or more partners. The SOI-R assesses attitudes about casual sex with three items using a 9-point Likert-type scale (1 = Strongly disagree; 9 = Strongly agree). The desire sub-facet assesses the frequency of experiencing sexual desire/fantasy with three items rated on a 9-point ordinal scale ranging from Never to At least once per day. We averaged items across the three subscales (behavior, attitude, and desire), with scores indicating an overall orientation toward casual sex, which had a sample mean of 4.24 (SD = 1.56; α = .82).
Procedures
IRB approval was obtained to collect online data from three recruitment sources. Most participants (n = 325) were recruited through Qualtrics, which prescreens and recruits participants with the help of partnered research panels. Qualtrics participants are identifiable to the panel that recruited them, but their identity was shielded from researchers. Though these participants cost $6 on average, Qualtrics provides variable compensation according to each individual’s agreement, and presumably claims some portion of the cost for themselves. To balance out idiosyncrasies tied to this survey platform (see Heen et al., 2014), we recruited additional participants through Amazon’s Mechanical Turk (MTurk; n = 20), a crowdsourcing marketplace for completing online tasks, as well as the experimental credit system for students enrolled in psychology courses at our mid-size Northeastern University (Sona Systems; n = 75). Sona participants received credit towards their psychology course experiment participation requirement. MTurk participants received $0.75. Funding was drawn from the senior author’s faculty startup funds. Although Sona participants cannot be directly linked to their online survey data, their participation credit must be granted manually and therefore the identities of those individuals who completed the study were viewed by at least one member of the research team.
Participants completed all measures online and were assured of anonymity. Participants were at least 18 years of age and gave informed consent. The order of the measures and tasks was as follows: demographic and social desirability measures, six vignettes with outcome questions, and then the personality and attitudinal measures in a randomized order. Vignettes were presented prior to the majority of surveys to prevent the latter from biasing responses to the former (see Lofgreen et al., 2017). The order of measures and tasks and the approach to vignette presentation did not differ across subsamples recruited from different sources. Average completion time for the full procedure was approximately 51 minutes (SD = 47).
Each of the six vignettes asked a participant to imagine himself in a hypothetical dating scenario with a woman he finds attractive and with whom he hopes to have sex. Each vignette culminated in the man attempting to escalate sexual activity, to which the woman either refuses, consents, or responds passively. Men received two vignettes per each type of communication, which – due to programming needs – were administered consecutively in blocks. The order of the blocks was randomized and each participant received each block only once. This manuscript focuses on the refusal vignettes, which had three variations of the female character's behavior: (1) nonverbal refusal: “pushes away from you”; (2) verbal refusal: “says ‘Let’s not do this right now’”; and (3) combined refusal: “pushes away from you and says ‘Let’s not do this right now’”. Preliminary models revealed that different communication types (verbal vs. nonverbal) did not differentially impact men’s sexual decision making, so these conditions were collapsed. The vignettes varied randomly on all the other situational parameters, allowing for a total 256 possible combinations (2 female attire x 4 female sexual history x 2 relationship sexual history x 4 alcohol consumption x 4 intimacy attained). The levels for each parameter were adapted from Lofgreen et al.’s (2017) work, though we uncoupled male and female alcohol use and added an additional level to the intimacy attained factor. See the Supplemental Appendix for examples of complete vignettes.
After reading each vignette, participants rated the likelihood that they would respond given eight possible behaviors on a 7-point Likert-type scale (0 = Very Unlikely, 6 = Very Likely). These behaviors were categorized as coercive (e.g., verbal persuasion) and forcible (e.g., ignoring the woman’s refusal and proceeding). These behaviors were presented in a consistent order, ranging from the most normatively and ethically acceptable response to a refusal (i.e., stopping the sexual encounter, which was reverse scored) to the most unacceptable (i.e., use of physical force). Omnibus scores were created by averaging across items within subscales. Scores on each index had a possible range of 0–6, with higher scores indicating a greater likelihood of using some coercive or forcible rape tactic. Coercion and force had respective sample means of 1.91 (SD = 1.51) and 1.29 (SD = 1.80), with α equal to .82 and .94, respectively. 1
Planned Analyses
We used SPSS 28 to explore the associations between our predictors and the coercion and force proclivity measures in separate random intercept models. The first level modeled the extent to which each individual’s responses varied between the two refusal vignettes. Characteristics of the vignettes were effect coded and entered in to explain variation at this level. These parameters were keyed so that a −1 indicated a level that is less likely to heighten consent perceptions (e.g., conservative attire), whereas a 1 represents a level more likely to be interpreted by men as indicating sexual interest or availability (e.g., revealing attire). Parameters with more than two levels were numerically scaled so that the levels summed to zero (e.g., −2, −1, 1, 2 for intimacy attained). This coding scheme renders the model intercept as the grand mean for each outcome collapsed across all situational variances, with the estimate for each situation parameter being the expected change in sexual assault proclivity associated with a change in that variable.
The second level modeled the random effects, with each participant having a separate intercept that could vary randomly from the grand mean for each outcome. Significant variation in the random intercepts implies that some men endorsed proclivities towards coercion and/or force more or less strongly than others across the two vignettes. Dispositional predictors were entered into the model to explain variation at this level. We first ran a model with no predictors to obtain baseline estimates for the variance components (i.e., random and residual) and information criteria for the model overall. We then entered the situational and dispositional variables into the model simultaneously. Person-level variables were standardized to facilitate interpretation and comparison of the regression weights. Cohen’s d was used as a measure of effect size for the situational factors, whereas r was used for the person-level covariates. We evaluated whether or not these variables improved model fit using a -2Log Likelihood (LL) test and provided estimates of explained variance at each level using a Pseudo-R2 metric. 2
Results
Descriptive Statistics and Regression Diagnostics
The modal amount of missing data per survey was zero, followed by 1 (except for the SOI-R). The number of cases per survey with some missing data ranged from 5.2% (n = 22) on the SOI-R to 24.8% (n = 104) on the LSRP. Missing item data were imputed using mean substitution except for cases missing more than half their responses on a given survey, 3 with 37 participants (9.0%) being designated as missing at least one scale score. For this subset of participants, the modal number of missing surveys was one (n = 14). Scale scores for these missing surveys were imputed using a regression-based strategy whereby the participants’ omnibus scores on the remaining indices were used to estimate the missing survey value. Due to a programming error, 17.6% of men lacked usable data for the IRI. This error only impacted participants from the MTurk (n = 16) and Sona samples (n = 41). The same imputation strategy was employed for these cases.
Study variables met the assumption of normality. Preliminary runs of the models identified several high influence cases using Cook’s D (Cook, 1977). Cook’s D is calculated for each individual case by comparing the change in predicted values across cases when omitting that case from the model. Values above the cutoff (4/N) suggest that inclusion of that case has a significant impact on the regression weights in the overall model. These analyses revealed that approximately 14.5% (n = 61) of men emerged as having a significantly disproportionate impact on the regression weights. Analyses revealed that the group of high influence cases scored higher than average in rape myth acceptance, t (75.90) = 1.77, p = .040, d = .27 (unequal variances); and endorsed greater benevolent sexism, t (418) = 3.30, p < .001, d = .46, right-wing authoritarianism, t (418) = 2.08, p = .019, d = .29, and hypermasculinity beliefs, t (418) = 1.84, p < .033, d = .26, as measured by the MRAS. Also, the high influence group demonstrated greater intention to use coercion, t (838) = 5.01, p < .001, d = .55, as well as greater force proclivity, t (838) = 2.22, p = .013, d = .22, with these effects being in the large and medium range, respectively.
Considered as a set, these findings align with the notion that there is a subset of “high risk” men who are different both characterologically and with respect to their patterns of sexual offending (Swartout et al., 2015a). That their presence in the sample has undue influence on the observed estimates could suggest that any effects are more reflective of these men than the broader sample. Given our present focus on the average emerging adult man’s sexual assault proclivity, we first tested our proposed models using the full sample, as intended, and then compared these results to those obtained when excluding highly influential cases. If influential cases are driving the results in the overall sample, then any observed effects should become null and significantly different when they are excluded. We also ran models exploring whether highly influential cases responded differently to situational cues than did the other men.
Primary Analyses
Regression Models Depicting Predictors of Coercion and Force.
Note. Multilevel multivariate models predicting use of coercive and forcible rape tactics. Initial models are presented first, followed by the models excluding the unduly influential cases, followed by the cross-level interaction effects between unduly influential individuals and the situational variables.
*p < .05. **p < .01. ***p < .001.
Intercept values were significantly different from zero for both coercion, t (422.74) = 36.84, d = 3.58, and force, t (421.79) = 20.72, d = 2.01, ps < .001; suggesting that men on average do not rule out the use of sexual assault tactics in response to a refusal. As hypothesized, higher scores on a subset of person-level variables uniquely corresponded to greater sexual assault proclivities. Specifically, a greater likelihood of coercive tactics was found for those with stronger endorsement of hypermasculine beliefs (AMIRS), t (421.84) = 3.31, p < .001, r = .16, psychopathic traits, t (421.67) = 2.99, p = .003, r = .14, and rape myth acceptance, t (423.33) = 8.65, p < .001, r = .39. Hypermasculine beliefs, t (421.19) = 4.01, p < .001, r = .19, psychopathic traits, t (421.104) = 2.76, p = .006, r = .13, and rape myth acceptance, t (422.17) = 11.09, p < .001, r = .47, also associated with a greater likelihood of the use of force. Hostile sexism also predicted the use of force, t (421.32) = −4.17, p < .001, r = .20, but the direction of this association indicates that it was a suppressor effect. Of the situational factors, the level of intimacy attained increased the likelihood that the man would use coercive tactics, t (693.102) = 4.77, p < .001, d = .36. Addition of the dispositional and situational variables as a set significantly improved both the coercion and force models, χ2 (19) = 291.2 and 357.9, respectively, ps < .001. The set of predictors explained 5.5% and 58.7% of the variance at the situation- and person-level, respectively, for coercion, as well as 1.2% and 61.7% of the variance in force, also respectively.
Ad-Hoc and Exploratory Models
Regression weights and standard errors for these models are presented alongside our primary analysis in Table 1 for comparison. Our ad-hoc analyses provided minimal support that the effects obtained using the full sample results from the influence of a subset of high-risk men. There was a significant effect for empathy on coercion when removing high influence men from the sample, t (359.66) = −1.98, p = .048, r = .10, which also attenuated the effects of psychopathy on force, t (360.91) = 1.64, p = .102, r = .09; but neither these nor any of the remaining estimates statistically differed between the two models. We also ran exploratory cross-level interaction effects using the variable differentiating the high-influence cases from the remaining sample and the situational factors. These analyses were based on the premise that specific situational factors may influence different types of men in unique ways, and those identified as influential were elevated on several rape-associated traits and attitudes. These analyses revealed several effects supportive of this notion.
Specifically, we found that high-influence men responded differently than other men for coercion when women wore revealing attire, t (701.77) = −2.73, p = .006, d = .21, and when he used alcohol, t (612.31) = 3.16, p = .002, d = .26; with a significant interaction with female alcohol use emerging for force, t (471.17) = −2.56, p = .011, d = .24. Plots of these effects showed that men identified as highly influential were more likely to use coercion when the woman was conservatively dressed (Figure 1(a)) and when they themselves had consumed alcohol (Figure 1(b)), but less likely to use force when only the woman had used alcohol (Figure 2). Across situations, their reported likelihood of using assault tactics was higher than men at large, who did not appear to respond differently across levels of the situational factors. These findings indicate that the highly influential cases were not driving the effects found for the full sample, but that situational factors appear to affect these men’s sexual decisions differently than for other men. High- and Low-Risk Men’s Use of Coercion Based on the Woman’s Attire (Above) and Men’s Alcohol Use (Below). Note. This figure plots the interaction effect showing that men identified as statistically influential (and at higher risk for perpetration) were more likely to endorse coercive tactics when the female character in the vignette was conservatively dressed (above) or when their own character in the vignette had consumed alcohol (below). High- and Low-Risk Men’s Use of Force Based on Female Character’s Alcohol Use. Note. This figure plots the interaction effect showing that men identified as statistically influential (and at higher risk for perpetration) were less likely to endorse forcible tactics when the female character in the vignette had consumed alcohol.
Discussion
We evaluated a broad array of situational and dispositional variables that have hitherto been highlighted as likely associates of sexual assault, and examined their unique associations with emerging adult men’s decision to use coercive and forcible rape tactics during a prototypic hookup using vignettes wherein a woman refuses his sexual advance. Broadly, we found that (a) men on average did not entirely rule out the use of coercive or forcible tactics when women refused their sexual advance during a hookup; (b) elevated risk predominantly reflected personal, rather than situational factors, particularly the decision to use force; (c) our slate of dispositional predictors accounted for a large share of the variance in sexual decision making, with several variables making unique contributions; and (d) while men’s self-reported likelihood of using coercive or forcible rape tactics varied from situation to situation, only the level of attained intimacy reliably predicted these variances across men. In addition, we found that (e) a subgroup of cases with a profile of elevated dispositions reported higher sexual assault proclivity and were differentially sensitive to situational factors, but (f) our findings from the full sample appear representative of emerging adult men, as opposed to being driven by this smaller subset of cases.
The strongest unique dispositional predictor of assault proclivity was rape myth acceptance. Had we examined this variable alone, any effect of rape myth adherence could theoretically reflect any number of third variables. For instance, the notion that a woman’s refusal may signify token resistance is consistent with gender norms inherent to sexism and hypermasculinity, whereas those high in psychopathy may adopt rape myths to justify their misdeeds. However, we controlled for these and other candidate explanatory variables, suggesting that the effects for rape myth acceptance may genuinely reflect predominant cultural misunderstandings of what constitutes a sexual assault. This finding takes on particular relevance in the context of a refusal; that is, some men might not be interpreting a no at face value because they do not believe it to reflect the woman’s genuine intentions. This is encouraging insofar as misconceptions about rape may be more amenable to change than deep-rooted personality traits (e.g., psychopathy; Sewall & Olver, 2019). Though some evidence suggests that rape myths are changeable, these effects can be nuanced (e.g., Davis & Liddell, 2002) and potentially moderated by aspects of the intervention (see Hudspith et al., 2021). Considering the existence of effective interventions for rape myth acceptance in samples of emerging adult men (e.g., the Men’s Program; Stewart, 2014), we hope the current findings serve as a call for renewed interest in dissemination and implementation, as they suggest rape myths may be a critical target.
We also found that psychopathy and hypermasculinity associated uniquely with sexual assault proclivity. While there have been stable associations between these constructs and rape myth acceptance (e.g., Mouilso & Calhoun, 2012), it is noteworthy that they each seem to contribute in unique ways, calling attention to different mechanisms. While a highly psychopathic man might be more willing to violate another’s rights for immediate gratification, this may operate independently of hypermasculinity, which could increase his motivation to undertake sexual conquest for his own sense of self-worth or peer approval. However, the exact mechanisms in either case remain unclear because these constructs further decompose into related, but discrete phenomena. With respect to psychopathy, some operations contain subdomains that may be uniquely related to men’s response to a sexual refusal and require separate approaches to change (Muñoz et al., 2011). As such, the present effects support ongoing research examining narrower mechanisms of risk within psychopathy, as the effects for the broad trait were not better accounted for by other known covariates.
Hypermasculinity is also multifaceted and offers many potential explanations for how related assaults may unfold: Does the hypermasculine man who hates women perpetrate as an act of aggression, while the man who believes women should be sexual gatekeepers believe that her refusal is token? Our findings shed some light on these questions. We controlled for generalized hostility towards women, as well as traditional gender roles and their corollary rape myths, so it is unlikely that mechanisms residing in these domains can explain the present effect. The effect for hypermasculine traits also was only observed for the AMIRS while controlling for the MRAS. Given their respective foci, the effect for the AMIRS may be highlighting relational dominance needs, low self-esteem, and/or male role strain (Chu et al., 2005). Our findings thus suggest that hypermasculine men may be sexually transgressing as a form of aggression in response to a rejection or perceived injury to their sense of self (see Bushman et al., 2003).
We also found significant response variance across vignettes, suggesting that men in general are taking situational factors into account. However, the only consistent predictor was the level of intimacy attained in the current interaction. This echoes earlier research that a woman’s consent to less intimate forms of interaction (e.g., sexual touching) is almost automatically interpreted as consent for greater intimacy (Beres, 2014). Our findings uniquely contribute by indicating that similar decisional heuristics are operating even when refusal is directly communicated. This may reflect various rape myths in action, such as the belief that a woman’s “no” really means “yes”, or that as intimacy increases, a woman becomes obliged to gratify the man. In any case, our findings do not support the claim that the increased sexual excitement pursuant to greater intimacy overwhelms a man’s sense of control. Our effects were not found in vivo, suggesting that increased intimacy is somehow leading men to think differently about the situation, as opposed to being genuinely overwhelmed by it in the moment.
The detection of a group of men whose patterns of responding significantly impacted the regression models contributes to the debate in the field regarding whether a subsample of men differ qualitatively in their characteristics and proclivity for sexual transgression (e.g., Abbey & McAuslan, 2004). That is, we found that men carrying elevations across several rape-associated traits indeed scored higher on our indices of sexual assault proclivity. Our data do not speak to whether these men are responsible for a disproportionate number of assaults among emerging adults, as some evidence suggests (e.g., Swartout et al., 2015) , but suggest that they are probably not responsible for all sexual assaults. Our effects appeared to be representative of men in general, as opposed to signifying the influence of a few “bad apples” driving the overall results. Nevertheless, we found that researching these men separately was fruitful, as they responded to situational factors differently than other men. For instance, these men reported being more likely to use coercion when the woman was conservatively dressed. This may suggest that these men view revealing attire as a sign of sexual availability (Moor, 2010), and thus assume that a conservatively dressed woman is less sexually open and will require a degree of coercion for him to achieve sexual gratification. Similarly, it is possible that these men viewed a woman’s non-consumption of alcohol as an indicator of her personal preference towards more conservative behavior across a variety of situations. This would explain why we also found that these men to be likelier to employ coercive techniques in situations wherein only they had consumed alcohol.
An alternative explanation for why these men were more coercive when only they consumed alcohol may reflect their concerns over third-party blame and accountability. That is, these men may have been likelier to use coercive techniques after consuming alcohol because they believe it would exonerate them from blame. This may explain our finding that these men were at an elevated risk for the use of force across all combinations of male and female alcohol use except one, which is when he had not used alcohol but his female partner did. Possibly, these men were viewing this situation as particularly risky for them with respect to third-party blame, fearing that use of force would accurately be perceived as him deliberately taking advantage of a vulnerable woman. In any case, these effects suggest that a subset of men are at an increased risk of sexual assault when alcohol is situationally present, irrespective of any physiological effect of alcohol on behavior at the time. Considered broadly, these exploratory findings coincide with research highlighting that subtypes of high risk men exist (e.g., Logan-Greene & Davis, 2011), and further suggest that these men may approach situational factors from a more calculated or instrumental perspective when deciding how to respond to a sexual refusal.
Methodological Considerations and Future Directions
Our study focused on men’s behavior following a woman’s sexual refusal in an initially consensual interaction. Though this scenario is prototypic in emerging adult populations, sexual assaults are not confined only to violations of the “no means no” criterion. Whether our current findings generalize to misconduct in other forms and contexts is uncertain. Likewise, our use of a heteronormative sexual script limits our abilities to draw conclusions about same-sex encounters and victimization of men or gender diverse individuals. A strength of the vignette design, however, is that it allowed for experimental control over other situational details, and for us to ethically explore men’s sexual decisions in the moments leading up to a potential assault. Whereas retrospective designs are subject to biased reporting (e.g., Zhang et al., 2018), the current design may provide some of the clearest insights into men’s decision making at a critical moment within a high-risk sexual interaction. Although we cannot be certain that the currently observed effects would match what happens in vivo, this approach also has an interpretational advantage: Physiological explanations for the role of particular factors (e.g., alcohol, physiological arousal) on sexual decision making can be ruled out, suggesting that the observed effects more likely relate purely to cognitive-affective phenomenon.
Our sexual assault proclivity measure also had strengths; most notably, the use of specific behavioral terms to characterize both parties’ actions instead of ambiguous descriptions or terminology that is value-laden (Koss et al., 2007). However, our measure sampled only a subset of tactics used to coerce sex from unwilling partners (cf., threats to withdraw affection; Pugh & Becker, 2018). Future operations could include additional persuasive tactics (e.g., threats to break up or humiliate the partner, promises of behavior change), and should consider a broader range of strategies, such as trying to elicit sexual arousal and making verbal threats of force (Katz & Tirone, 2010). Fine-grained analyses could also shed light on rape tactics that begin prior to actual sexual interactions (e.g., getting a date intoxicated; Lorenz & Ullman, 2016).
It is also important to highlight that we sampled a population who are particularly likely to perpetrate sexual assaults but remain under-identified by victims and under-reported to authorities (Abbey et al., 2012). Although studies on individuals facing legal consequences are important, those findings may be fundamentally unrepresentative of sexual misconduct broadly given the low prosecution and conviction rates (National Research Council, 2014), as well as evidence that men who use certain rape tactics are less likely to face legal action (Lapsey et al., 2022). Studies of individuals who self-report undetected perpetration histories can address this limitation, but still exclude men who are at a high risk for rape but have not actually done so (e.g., those who thus far have had only consensual partners). As such, the current findings contribute knowledge pertaining to what situational factors incline a “typical” emerging adult man to become sexually assaultive, as well as which dispositional differences across men increase this likelihood. Regarding the situational influences, however, we ultimately found that most of the response variance tied to the vignettes was left unexplained, raising the question of what accounts for men’s differential responding across these scenarios if not their variable components? We speculate that the combinations of situational factors and complex person-by-situation interactions may be in play. Additional research is required to parse out which, if any, of these explanations helps account for how situational factors influence men’s proclivity towards coercive and forcible assault tactics.
Implications for Intervention
Several of our findings have implications for intervention efforts targeting sexual assault by emerging adult men. It is first important to highlight that our sample of men, on average, did not discount the use of coercive or forcible tactics when faced with a refusal to their sexual advance. Though the reported likelihoods were low, these findings should be considered in light of the fact that the women in the vignettes explicitly conveyed sexual refusal. Though some data suggests that increased assertiveness may help protect women against sexual victimization (Kelley et al., 2016), our findings suggest that this might not be entirely protective, at least within heterosexual hookup scenarios. This seems like important information to impart unto women through educational campaigns, or as part of mental health treatments for survivors blaming themselves for their own assault, a correlate of post-traumatic stress symptoms (Ullman et al., 2007). Moreover, these findings somewhat undercut existing sexual educational campaigns targeting consent communication that posit the use of clear affirmative indicators will help curtail sexual assault (see Mattson et al., 2022). This may be to some extent true, but the current findings suggest that many men view an unambiguous “no” as negotiable or irrelevant, which may require a different approach to prevention than what is offered by affirmative consent education.
Second, we found that attitude and personality traits accounted for most of the response variance in coercion and force, meaning that characteristics of the man may be the more relevant determinant of a sexual assault than the particulars of the sexual situation in which he finds himself. This is notable considering that situational factors are often used to blame victims of sexual assault by important third-parties (e.g., police; Venema, 2019), whereas our findings suggest that few situational features appear reliably associated with men’s sexual decisions during hookups. Rather, it is that certain men are likelier to use coercion and force across situations, and that a subsample of men elevated across particular traits were especially dangerous. These results coincide with Swartout et al.'s (2015a) findings that a relatively large subsample of college men are deliberate repeat offenders. Though there is room for educational interventions targeting particular attitudes (e.g., rape myth acceptance), it is possible that what is also needed are strategies that increase perpetrator accountability. That is, it is plausible that the high rates of sexual assaults are partially attributable to the low likelihood of them being reported, let alone convicted of a sexual offence (Morgan & Truman, 2020). Successful approaches might involve helping victims acknowledge and disclose a sexual assault (Edwards et al., 2022), as well as increase the likelihood of legal consequences for offenders (e.g., improved police interviewing practices; McQueen & Murphy-Oikonen, 2023).
Third, men’s reported likelihood of using coercion or force increased at higher levels of intimacy attained in the interaction. These findings could reflect that increases in the level of intimacy more strongly activated emerging adult men’s mental script for scenarios that end with consensual sex. This may have then shaped how they processed subsequent communication, for instance, rendering them more likely to interpret refusal as token resistance. It is alternatively possible that higher levels of intimacy make men more confident that coercive tactics would be effective, or that forcible tactics would be ignored or excused given extant rape myths about women’s obligations to gratify men who have become sexually aroused. Without knowing the exact mechanism, however, incorporating this finding into existing prevention or intervention efforts is currently premature. But the increased likelihood for sexual misconduct at higher levels of intimacy appears to be a robust effect (e.g., Lofgreen et al., 2017). As such, additional research explaining this phenomenon has clear applications for intervention.
Conclusion
We found that variance in sexual assault proclivity largely exists at the person-level; and that rape myth acceptance, hypermasculinity, and psychopathy are uniquely relevant to explaining decisions made by emerging adult men. The level of intimacy attained contributed small, albeit stable effects across situations; whereas other elements of the situation had more nuanced influences that took into account profile-level elevations across traits related to assault. More specifically, our findings provide further evidence that a high-risk group of men exists, and showed that they respond differently to situational factors associated with sexual assault. Additional work focusing on men’s assault proclivity and perpetration should consider how these findings may be conditioned on these profile-level elevations. Future studies should also incorporate additional person-level predictors, separate omnibus measures by sub-constructs, and explore person-by-situation as well as situation-by-situation interactions. Such findings have the potential to identify the most relevant specific mechanisms underlying a man’s decision to use coercive or forcible rape tactics, thus providing critical information to devise and refine interventions that target the widespread problem of sexual assault.
Supplemental Material
Supplemental Material - Does No Mean No? Situational and Dispositional Factors Influence Emerging Adult Men’s Intentions to Use Assault Tactics in Response to Women’s Sexual Refusal During Hookups
Supplemental Material for Does No Mean No? Situational and Dispositional Factors Influence Emerging Adult Men’s Intentions to Use Assault Tactics in Response to Women’s Sexual Refusal During Hookups by Allison M. McKinnon, Richard E. Mattson, and Ashton M. Lofgreen in Sexual Abuse.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Supplemental Material
Supplemental material for this article is available online.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
